Here is the entire strategy in one sentence:
"Today, which has stronger 30-day momentum — BTC, ETH, or SOL? Hold the strongest positive one. If all three are negative, hold cash."
That's it. Three assets, one comparison, run once per day.
Over 5.7 years (2020-08-11 to 2026-04-17), this rule returned +34,929% while BTC HODL returned approximately +901% over the same period. Yes, that number is real. It's also driven heavily by SOL's 2020-2021 explosion, so let's break it down honestly below.
Walk-forward: 2 of 3 windows beat BTC HODL. Parameter robust across every lookback from 10 to 30 days (clear plateau, not a single-peak overfit). Paper-traded since 2026-04-17. The Tri-Rotator is our seventh live bot.
Why Add a Third Asset
The original Rotator (BTC/ETH only, 21d momentum) already passes Walk-Forward 3/3 with +3,020% over 8 years. So why a 3-asset version?
Because crypto regimes shift between MORE than two assets.
Every cycle has its narrative leaders:
- 2017: ETH (ICOs)
- 2020-2021: SOL (DeFi summer + meme rotation)
- 2024-2025: SOL again (memecoin season)
- 2025-2026: BTC (institutional ETF era)
A 2-asset bot misses SOL outperformance entirely. The 3-asset version captures it when it happens, falls back to BTC/ETH otherwise.
Why no LTC, BNB, XRP, etc.? Three reasons:
- SOL has enough liquidity and volume to be tradeable for a $10k portfolio
- SOL has a unique narrative (high-throughput L1) that genuinely diverges from BTC/ETH
- Adding more assets dilutes the signal — at 5+ assets you're essentially holding a basket
Three is the sweet spot tested.
The Backtest, With Caveats
| Metric | Tri-Rotator (M-30) | Rotator (M-21 BTC/ETH) | BTC HODL |
|---|---|---|---|
| Period tested | 2020-08 → 2026-04 (5.7y) | 2018-01 → 2026-04 (8y) | varies |
| Total return | +34,929% | +3,020% | +901% (5.7y window) |
| Walk-Forward | 2/3 windows beat HODL | 3/3 | n/a |
| Trades/week | 0.95 | 0.91 | 0.005 |
| Max drawdown | -57% | -72% | -77% |
The +34,929% needs context. Most of it comes from SOL's 2020-2021 run, which took SOL from $1 to $260 (a ~260× move). The Tri-Rotator caught significant chunks of that. BTC HODL alone over the same window was +901%. Even ETH HODL was +1,041%. The 3-way rotation captured the SOL premium that single-asset HODL missed entirely.
Honest read: the 5.7-year backtest is shorter than the 8-year Rotator backtest because SOL only became liquid on Binance in August 2020. Future cycles may not have a similar SOL-style moonshot. The strategy CONCEPT (rotate to strongest momentum) is sound; the SPECIFIC return number depends on whether future regimes have similar relative-strength dynamics.
The Walk-Forward score is 2/3 (vs Rotator's 3/3). The third window where it loses is the recent strong BTC bull regime (2024-2026). Same structural limitation as every momentum strategy: when one asset is in violent uptrend, momentum strategies underperform straight HODL.
Parameter Robustness — The Plateau Test
After Sharpshooter's overfit lottery, every new strategy must pass the parameter-neighborhood sweep:
| Lookback | Full Return | Walk-Forward |
|---|---|---|
| M-10 | +3,709% | 2/3 ✓ |
| M-14 | +5,785% | 2/3 ✓ |
| M-18 | +2,760% | 2/3 ✓ |
| M-21 | +6,228% | 2/3 ✓ |
| M-25 | +20,466% | 2/3 ✓ |
| M-30 | +34,929% | 2/3 ✓ |
Every lookback in the sweep passes 2/3 walk-forward. This is the opposite of Sharpshooter (M-14 was a single isolated peak with neighbors at 1/3). A strong plateau across all tested lookbacks means the strategy class has real signal — the specific lookback choice is fine-tuning, not lottery.
We picked M-30 because:
- Highest absolute return (within the plateau)
- Lower trading frequency = lower fee drag
- Filters out short-term noise that can flip BTC↔ETH↔SOL signals weekly
Pickng M-21 instead would also work (+6,228%, similar WF). The plateau means you can't go very wrong within the band.
How It Compares to The Rotator
The Rotator (BTC/ETH 21d) and the Tri-Rotator (BTC/ETH/SOL 30d) overlap conceptually. Both are momentum-rotation bots. Why run both?
| Bot | Universe | Lookback | Period | Pass criterion |
|---|---|---|---|---|
| Rotator | 2 assets (BTC, ETH) | 21d | 8 years | WF 3/3, plateau |
| Tri-Rotator | 3 assets (BTC, ETH, SOL) | 30d | 5.7 years | WF 2/3, plateau |
Pro-Rotator (BTC/ETH only): longer test window (8y), tighter WF result (3/3), lower drawdown profile.
Pro-Tri-Rotator (BTC/ETH/SOL): captures additional asset class, dramatically higher backtest return (because SOL was wild in the test window), demonstrates that the rotation concept generalizes beyond 2 assets.
Decision: keep both. They will probably correlate in BTC-led regimes (both buy BTC) and diverge during SOL/altcoin seasons (Tri-Rotator captures SOL, Rotator stays on BTC). After 90+ days of live data, we'll know whether the extra SOL exposure adds Sharpe or just adds variance.
If after 90 days Tri-Rotator's SOL trades systematically degrade live performance, we'll retire it. If they add value, Rotator might get retired in favor of Tri-Rotator. Honest decision driven by data.
The First Live Trade
The bot was initialized at 2026-04-17 18:39 UTC. First check immediately fired:
- BTC 30d momentum: +8.54%
- ETH 30d momentum: +10.15% ← winner
- SOL 30d momentum: -0.98% ← skipped (negative)
Signal: ETH. Bought 4.12 ETH at $2,425. Portfolio: $9,990 after $10 fee.
The fact that SOL was skipped on day one demonstrates the strategy working as designed: when an asset isn't trending positive, we don't buy it. We don't FOMO into "well, two are positive, let me also add SOL anyway" — we hold ONLY the strongest positive asset, full position.
Daily Mechanics
The bot runs four times per day via cron (00:15, 06:15, 12:15, 18:15 local). Each run:
- Fetches latest 60 daily bars for BTCUSDT, ETHUSDT, SOLUSDT from Bybit
- Computes 30-day momentum:
(today / 30-days-ago) - 1for each - Picks signal: strongest positive, or CASH if all negative
- If signal differs from current position: liquidate current to cash, then buy target. 2 trades per rotation, 0.10% fee each.
- If signal equals current: do nothing (idempotent)
- Log to CSV, save state, send Telegram only on rotation events
You see every rotation via Telegram or email (newsletter). You decide if you want to follow.
Honest Caveats
Single-cycle test. 5.7 years is one full BTC cycle including the 2020-2021 mega-bull, the 2022 bear, and the 2023-2026 recovery. Future cycles with different relative-strength dynamics may produce different results.
SOL-specific risk. SOL has had three significant outage events in its history (2021, 2022, 2023). If SOL has a major outage during a rotation, the bot may be forced to hold a position that it can't exit cleanly. Backtest doesn't model this.
Tax classification (Switzerland). ~50 trades per year is borderline professional trading territory. Real-money deployment would need a tax review.
Cross-exchange basis risk. The bot uses Bybit prices for SOL/ETH/BTC. Real execution would happen at Bybit's spot prices. If you replicate on a different exchange, basis differences could affect returns.
The +34,929% is window-dependent. Replicating this exact strategy in a different 5-year window (2015-2020 for example, if data existed) would produce a very different number. The strategy CLASS has edge; the specific magnitude is sample-specific.
What Happens From Here
- Watch: /bots page Live Performance Card updates after each rotation. Daily log in
/110-BotLab/tri-rotator/state/daily_log.csv. - 90-day review: compare Tri-Rotator vs Rotator vs Tactician 2.0 live performance. Decide retirements based on data.
- Telegram: every ROTATE event sent to subscribers in real time.
Either of two outcomes after 90 days will be published as a follow-up post:
- "Tri-Rotator validated — Rotator retired" (if SOL exposure adds value)
- "Tri-Rotator failed live — backtest was misleading" (if it doesn't)
Related reading:
- The Rotator — the BTC/ETH-only sibling
- The Tactician — single-asset BTC momentum (now Risk-Parity edition)
- Beat HODL or Don't Bother — the benchmark every strategy must clear
- Parameter Robustness — why we sweep neighborhoods
- Sharpshooter Post-Mortem — what an isolated peak looks like (vs Tri-Rotator's plateau)
- HF Alpha Hunt v4 — full methodology + the 3 angles that died
This bot in the post-mortem ledger: See /post-mortems →
Every retired strategy and failed walk-forward — documented publicly.
